1. Field of the Invention
The invention relates generally to the field of digital data processing systems and more specifically to local area networks in which a plurality of such systems are interconnected to provide distributed processing capabilities to a number of users. In particular the invention provides improvements in message transfer protocols for local area networks to enhance the message transfer capability of the network.
2. Description of the Prior Art
As small and medium-sized computer systems are becoming less expensive and more powerful, a number of them are being interconnected to form networks to ensure that a number of different types of services are available at any time to users having diverse processing needs. Such services may include any of the services which are normally available from such networks, including such as electronic mail (mail storage and forwarding), word processing, accounting, such as payroll or inventory, or data communications over telephone lines or microwave links. Interconnecting systems into a network helps to enhance the availability of services to service users by including a number of service providers in the network and having each provider provide one or more of the services, thus reducing the likelihood ,that the failure of any one service provider in the network will result in a significant number of services being unavailable to the users at any one time. Indeed, a local network may be arranged so as to have several service providers providing overlapping services, in which case several providers have the ability to provide a particular service if one service provider fails or is saturated with service requests.
Typically in a local area network, the communications in the network take place over one or a limited number of communications links. Examples of such communications links include those defined by well-known DECnet, SNA (System Network Architecture) or X.25 communications protocols using data links such as Ethernet. A number of service users, such as, for example, video terminals controlled by operators, are connected through interface devices known as "terminal servers" to the communications link. Similarly, the service providers are connected to the communications link through interface devices known as "nodes".
If the operator desires to use a service provided by a unit connected to a node, it may request connection to the node and, through the node, to the unit to have the service provided. Normally, the operator has to know the particular node(s) and unit(s) that provides the desired service. The operator selects a node and unit to provide the service, and causes the terminal server to request service by that node and service provider. The terminal server and the node exchange messages which enable a "virtual circuit" to be established which provides a data transfer mechanism between the operator's terminal and the provider of the service. The virtual circuit essentially extends from the operator's terminal, as the service user, through the terminal server, over the communications link and through the node to the service provider. If a number of users are using the local area network, several virtual circuits may be established over the communications link to provide communications between the users and providers. In addition, if several terminals connected to one terminal server require services from a service provider connected to the same node, separate virtual circuits are normally established between each terminal and service provider providing the required service.
Service data is transmitted in the form of messages through the virtual circuits between the terminals and the service providers. All of the messages are queued by the terminal server and transmitted over the single communications link. To ensure that the messages are received, units connected in the network re-transmit the message until an acknowledgement from the recipient is received verifying correct receipt. Specifically, the terminal servers and nodes, after they transmit messages through the virtual circuits over the communications links, monitor the communications link for acknowledgements and if no acknowledgement is received within a selected period of time, corrective action is taken. Each message transmitted through a virtual circuit is acknowledged by a separate acknowledgement message through the virtual circuit, even if a series of messages are transmitted between the same terminal server and node through different virtual circuits over the same communications link. Each acknowledgement message must be separately generated, and thus requires that time and facilities be dedicated at the receiving device to the generation of the separate acknowledgement messages. Furthermore, requiring such separate acknowledgement messages could cause the communications link to quickly become unnecessarily burdened.
In current networks, message transfers through virtual circuits over a communications link are initiated either by the occurrence of certain events ("event driven" transmission), such as the presence of data to be transmitted, or by the timing-out of certain timers ("timer-based" transmission). Both the event-driven and timer based message transfer systems incorporate certain assumptions about message traffic through virtual circuits over the communication link. The event driven systems assume that the communications link has sufficient bandwidth, even when it is being heavily used, to ensure that messages can be delivered from the terminal servers to the nodes, the messages can be processed by the service providers, and the responses to the messages can be returned to the terminal servers, all within a maximum tolerable delay period. If the communications link carries too high a level of message traffic, the delays will, however, become unacceptably long. Furthermore, if a unit transmits data every time it receives a few bytes of data from a service user or provider, a significant number of such messages will be composed primarily of virtual circuit identification information, which is necessary to ensure proper identification of the virtual circuit carrying message over the multiplexed communications link, but otherwise serves no purpose.
Timer-based message transmission systems ensure, on the other hand, that every unit connected to the communications link will be able to transmit messages over the link at periodic intervals. These systems ensure that all of the units have relatively uniform access to the communications link. However, such systems also have a number of deficiencies. First, each unit, when its time comes to transmit, transmits messages through its virtual circuits whether or not it has any data to transmit, obviously wasting bandwidth on the communications link. Furthermore, as units are added to the system, the timers of all of the units would have to be adjusted to ensure that all of the units have reasonably equal access to the communications link.